Scent Dog ID: Science or Pseudoscience?

Always Faithful, Doberman, Military Working Do...
Image by BL1961 via Flickr

I’ve blogged in the past about “Man’s Best Friend” and the archaeological evidence of the domestication of the human by canines (or perhaps I have that reversed). I’m a big fan of dogs and what I see as the special relationship humans have with their canine friends. By and large, it’s always been a symbiotic relationship where the dogs earn food, shelter and protection from the wild in exchange for hunting skills, sounding the alarm on intruders, etc. Both of us get friendship and companionship from the deal -we become members of each other’s families in spite of the difference in species.

I’m also a big fan of working and service dogs that provide guidance to the blind, remain alert for the onset of seizures, search for lost people, locate bombs and drugs, support police departments in apprehension of criminals, and so on.

Don’t end up hounded by a police dog, go to Drug Rehab (¬†and get help.

But, here in my fair state of Texas there is a group working to overturn the convictions and use of working dogs that provide “scent identification” of criminals. The way this is done is that a line up is created by swabbing the suspect(s) and several controls, placing the swabs in separate cans, then giving the dog a scent from the crime scene to identify.

At first, this seemed like a totally awesome and rational method of identifying suspects! I immediately recalled the time when my wife and I went to visit my mother at her apartment for the first time, taking with us our Brown Labrador (named Coffee). Neither the dog nor my wife had ever been to the apartment, the door to which was nestled approximately in the middle among half a dozen others. While I was grabbing our overnight bag from the car, the dog was let out. When I looked up, she was waiting at the correct door to my mother’s apartment! She could smell the right choice.

But this anecdote in no way implies that scent dog identification is a realistic or practical alternative to other forms of evidence in convicting people of crimes -particularly when so much is at stake in the way of lives and justice. Among the criticisms leveled by the Innocence Project of Texas is that there are no proper methodologies for use and employment of canines in scent identification and that the chance for error is so great as to render the idea completely unreliable. I heard this story on the local NPR station yesterday and thought right away that there is potentially a simple way to answer this issue: review the scientific literature on the subject.

And here’s what I found:

The Genetic and Environmental Variable

I first wanted to know what was understood, scientifically, about human scent and the canine interpretation of it so I read Harvey et al. (2006) ((Harvey, Lisa M.; et al. (2006). The Use of Bloodhounds in Determining the Impact of Genetics and the Environment on the Expression of Human Odortype
Journal of Forensic Sciences, 51(5), 1109-1114)) in which the authors study “the use of bloodhounds in determining the impact of genetics and the environment on the expression of human odortype.” Their interest was actually on the use of bloodhounds to track fleeing felons and missing persons. What they wanted to know was this: could a dog tell the difference between people who were related and would it make a difference if these relatives lived together or not?

What they found was that genetic cues weighed heavily on the dog’s ability to get a scent of an individual whereas environmental cues didn’t appear to significantly help the dog find a scent. The dogs were generally able to tell the difference between relatives even if they lived together except for when it came to monozygotic twins (siblings who come from the same egg). The hounds couldn’t do better than chance in discriminating between the two. In this instance, the dogs were mostly correct -sometimes 100% of the time. Good news for scent identification? Not necessarily.

Harvey et al. clearly show the application for tracking a missing person. In these tests, it was either that the dog had the right trail or it didn’t. In the case of scent identification, there is the issue of a dog trying to pick out the scent of a person who may or may not match the control scent. The obvious question is this: if a dog is presented with a set of scents in which none match the crime scene scent, what does the dog do? Will it alert on the closest match or go away without a reward? Is the dog expecting a biscuit, a favorite toy, or pat on the head if it does well? Havey et al. offer the following in there conclusion, which they cite from the Jury Rules of a 1983 court case:

Dog tracking evidence is not by itself sufficient to permit an inference that the defendant is guilty of the crime. Before guilt may be inferred, there must be other evidence that supports the accuracy of the identification of the defendant as the perpetrator of the crime

What About Scent Lineups?

Two of the lead authors of this study have concluded in a related study that dogs can be as much as 77.5% effective in following a trail through even an urban environment up to 48 hours after it was created ((Harvey, L.M. and Harvey, J.W. (2003). Reliability of bloodhounds in criminal investigations. Journal of Forensic Sciences 48, pp. 811-816)).

Their conclusion in that study included this statement:

Our experiments show that while dogs are capable of performing scent lineups, in the simple experimental setup of a choice between six, a large number of mistakes are made. However, the new identification task and the yearly examination improve on this, therefore the level of performance found in this study cannot be extrapolated.

They found that differences between dogs and trial types existed, calling into question any sort of reliable replication of methodology. Each dog is an individual and the training each obtains is often very different. Finally, they found that familiarity with the scents significantly influenced results. The identification “hits” were as much as 73% in some trials, but these were cases where the “suspect” scent was already known to the dog. In cases where the scent was familiar from training, the hits were as high as 67%. But in cases where the dog was unfamiliar with the scent (i.e. a complete stranger’s scent), hits were only 25%.

What Harvey and Harvey show is that there is some potential with the use of dogs for scent identification, but it requires additional study and the establishment of consistency in training and methods as well as careful evaluation of the individual dogs.

Based on the results of these two studies, I see know reason why dog scent identification, in its current use, should be any more reliable than dowsing. And even once training and methods are standardized, its clear that a scent identification should not be the primary evidence against a suspect. Rather it should be the lead used to help forensic scientists narrow their investigations.

Enhanced by Zemanta
About Carl Feagans 394 Articles
Professional archaeologist that currently works for the United States Forest Service at the Land Between the Lakes Recreation Area in Kentucky and Tennessee. I'm also a 12-year veteran of the U.S. Army and spent another 10 years doing adventure programming with at-risk teens before earning my master's degree at the University of Texas at Arlington.